A Sexually Dimorphic Learned Birdsong in the Northern Cardinal ’

The Condor 100:504-5 I I 0 The Cooper Ornithological Sowsty 1998

A SEXUALLY DIMORPHIC LEARNED BIRDSONG IN THE NORTHERN CARDINAL ’

AYAKO YAMAGUCH? Animal Communication Laboratory and Animal Behavior Graduate Group, University of California, Davis, CA 95616

Abstract In most temperatezone songbirds,singing is solely a behavior of males. How- ever, in some species, such as the Northern Cardinal (Cardinalis cardinalis), both sexes sing, and in both, songs are learned. The song of male and female cardinals soundsvery similar to the human ear. Whether songslearned by the two sexes can be sexually dimorphic in acoustic structurehas never been examined. In the present study, I carried out detailed analysesof the acousticstructure of male and female songs. Songs of the two sexes proved to be dimorphic with respect to the degree of syllable stereotypy and the amplitude of harmonics. These findings show that vocalizationsthat are learned by the two sexes can be sexually dimorphic. Key words: bioacoustics, birdsong Cardinalis cardinalis, sexual dimorphism, vocal learning.

INTRODUCTION portant implications for sexual dimorphism. Un- In many songbirds, singing is a behavior of learned vocalizations develop under genetic con- males. However, in some species, both sexes trol independently of auditory experience (Kon- sing learned songs during the breeding season. ishi 1963). In species with sexually dimorphic If both sexes produce songs, it is natural to ask calls, genetically determined vocal development whether the songs of males and females are dis- is likely to give rise to stereotypedvocalizations tinct so that they can be used for acoustic sex that are predictable and universal among mem- recognition. Although many bird vocalizations bers of each sex within a species, in much the have been shown to be sexually dimorphic same way as other sexually dimorphic traits, (Gahr and Gtittinger 1986, Taoka and Okumura such as plumage color, which develop relatively 1990, Ballintijn et al. 1997), whether songs of uniformly within each sex. In the case of learned oscines are sexually dimorphic has never been vocalizations, however, the extent of acoustic systematically examined. In this study, I exam- variability introduced in the process of vocal ined the acoustic morphology of male and fe- learning argues against the presence of consis- male Northern Cardinals (Cardinalis cardinalis), tent and universal sex-specificacoustic traits that a species in the Temperate Zone in which fe- can be reliably used for sex recognition. Can males sing as well as males. learned songs of oscines be sexually dimorphic, The vocalizations of birds can be generally and if so, how do they differ between the two separatedinto calls and songs. Songs of oscines sexes? differ from calls in one major way: learning usu- I have explored this issue in the Northern Car- ally plays more of a role in the development of dinal, a monogamous,nonmigratory speciesrang- song than of calls (reviewed by Kroodsma ing from south-eastern Canada to Guatemala 1982). All the vocalizations that are demonstrat- (Bent 1968). Both sexes of the Northern Cardi- ed to be sexually dimorphic are either catego- nal learn to sing during the first year of their rized as calls or it has yet to be shown that they lives (Lemon and Scott 1966, Dittus and Lemon are truly learned by both sexes. The develop- 1969, Yamaguchi 1996), and in both sexessongs mental differences of calls and songs have im- are used only during the breeding season (Las- key 1944, Bent 1968, Ritchison 1988). Male and female cardinals have individual repertoires of 8 I Received 15 December 1997. Accepted 17 March to 12 song types, most of which are shared 1998. among all individuals in a local population. ZCurrent address: Department of Biological Sci- ences, Sherman Fairchild Center for Life Sciences, There is no song type that is unique to one sex Columbia University, New York, NY 10027, e-mail: (Halkin 1990). Initially, I did not expect to find [email protected] any differences in the songs of the two sexes,

PO41 SEXUALLY DIMORPHIC LEARNED SONG 505 both becauseof the developmental plasticity dis- in the same male. Songs were digitized with a cussedabove, and because the songs of the two sample rate of 25K pts see-I, high-pass filtered sexessound very similar to the human ear. How- at 1.5 kHz and anti-aliasing filtered at 10 kHz. ever, field playback experiments show that cardinals respond differently to songs of the two FREQUENCY MEASUREMENTS sexes,suggesting that there are some differences Sound spectrograms were generated with a in the songs of the two sexes and that the birds frame length of 1,024 pts Discrete Fourier are capable of recognizing them (Yamaguchi, in Transform (frequency resolution = 24.4 Hz), a press). time step size of 200 (time grid resolution varied In the present experiment, I carried out de- between 6.5 msec and 24 msec), and minimum tailed sound analyses to characterize acoustic amplitude of -33 dB relative to peak. An on- parameters of the songs of the two sexes, and screen cursor was used to measure the maximum compared these between the two sexes. The re- and minimum frequencies of each song phrase. sults show that there are consistent and signifi- The frequency range of each song was computed cant differences in the acoustic structure of the from these measurements.An average dominant songs of the two sexes, despite their striking frequency for an individual song was calculated similarity to the human ear. based on the dominant frequencies of the first 1.31 set of each phrase from an amplitude spec- METHODS trum (32K pts spectrum range).

SAMPLE SONG RECORDINGS SYLLABLE REPETITION ACCURACY Eighteen male and 18 female songs recorded Cardinal songs typically consist of a few phras- from 11 males and 9 females in southern Ari- es, each phrase containing notes or syllables that zona between March and June of 1993 and 1994 are repeated multiple times. A note was defined were subjected to sound analyses. One to three as a continuous tracing on a spectrogram, sep- song exemplars were collected from each indi- arated from other tracings by more than 8 msec, vidual. All songs were recorded using a Dan and a syllable was defined as a group of notes Gibson parabola microphone and a Sony that is repeated in a phrase. In cardinal songs, TCDSM cassette tape deck. Great care was syllables and notes are typically repeated within taken to avoid overloading while recording. a phrase in a stereotypedmanner. To test wheth- There was no systematic difference in recording er the degree of syllable/note stereotypy within distance in males and females. Eight song types a song differs between the sexes, I computed were represented in the pool of 36 songs ana- pair-wise cross correlation coefficients of sylla- lyzed. Song types were shared by almost all bles (or notes) based on digital spectrograms. birds residing in one location, and were acous- Cross correlation is an objective way to evaluate tically distinct from each other. Acoustic varia- the structural similarity of a pair of spectro- tion found in different song types could con- grams; higher coefficients indicate greater simi- found the analyses of sex differences. To control larity between the two sounds (Clark et al. for this effect, a sample of songs was chosen 1987). Phrases consisting of repeated notes or with equal number of male and female exem- syllables were disassembledand stored as indi- plars (two to three from different individuals) of vidual notes. One to four consecutive pairs of each song type. syllables or notes from each phrase (e.g., Fig. The following six acoustic parameters were 1A) were used for cross correlation. The same measuredusing the SIGNAL sound analysis sys- number of syllable/note pairs were sampled to tem (Engineering Design 1987): (1) maximum calculate cross correlation coefficients for each frequency, (2) minimum frequency, (3) domi- pair of male and female songs. Spectrogramsof nant frequency, (4) frequency range, (5) syllable each note were generated with a frame length of repetition accuracy, and (6) the relative ampli- 256 pts and a time step size of 200. Average tude of harmonics. Song duration varies greatly, cross correlation coefficients were first calculat- making it an unlikely candidate for consistent ed for each phrase, then the average coefficient sexual dimorphism (Lemon and Chatfield 1971, acrossphrases within each song were computed Yamaguchi, pers. observ.). For example, song for further statistical analysis. I labeled the av- duration can vary between 0.40 to 5.52 set with- erage cross correlation coefficient of the entire 506 AYAKO YAMAGUCHI

song the “Syllable Repetition Accuracy” (SRA) tors). I first tested whether the variation in pa- of that song. For all sample songs of each song rameters is explained by sex and song types. If type, equal numbers of note pairs from similar the results indicated that sex significantly influ- positions within each song were used for SRA enced the measured parameters, then a post hoc calculation to control for sampling error. two-way Bonferroni test was carried out to test for the difference between the two sexes. If, in HARMONIC AMPLITUDE DIFFERENCE addition to sex, song type significantly influ- There are harmonic overtones in cardinal songs, enced the parameter, then I considered whether but in most of them, the fundamental frequency the interaction between sex and song type was is the loudest and the amplitudes of harmonics significant. In other words, the interaction effect are considerably suppressed. To determine was examined to determine whether the patterns whether the degree of harmonic suppressionvar- of sex difference observed are universal across ied between the sexes,I quantified the amplitude all song types. Residuals were examined and it difference between the fundamental frequency was confirmed that the data did not deviate from and the second harmonic and compared the fig- the statistical assumptions.The significance lev- ures for the two sexes. Power spectra were cal- el of 0.05 was used for all statistical tests. culated based on three different time frames within each song (e.g., Fig. lA, shaded areas RESULTS marked with xf), with a spectrum window of FREQUENCY MEASURES 512 pts (time resolution of 20.5 msec, frequency As already indicated, there are no song-type dif- resolution of 48.8 Hz). Time frames with fun- ferences between the sexes;both sexesshare the damental frequencies lower than 5 kHz were same song types. None of the frequency param- randomly chosen for power spectra calculation, eters measured (maximum, minimum, dominant so that all the second harmonics to be analyzed frequency, and frequency range) differed be- fell within the 10 kHz spectrum analysis range, tween the songs of the two sexes (Table 1, Fig. which covers the hearing range of most passer- 2). Thus, songs of the two sexes do not differ ines (Fay 1988). Within all songs of a given with respect to frequency. song type, approximately homologous segments of syllables or notes were chosen to produce a SYLLABLE REPETITION ACCURACY (SRA) spectrum in order to minimize sampling error. Syllable stereotypy varied significantly between The amplitudes at the fundamental frequency the songs of the two sexes (Table l), but not and at the second harmonic were measured us- among different song types (F7,20 = 2.4, P = ing an on-screen cursor, and the difference be- 0.06). Males repeated their syllables with sig- tween the two was calculated (Fig. 1B). Ampli- nificantly greater stereotypy than females. There tude differences from three spectra were aver- was no significant interaction effect of sex and aged for a song, and I labeled this average the song type (&, = 0.4, P > 0.5) so that male “Harmonic Amplitude Difference” (HAD) of songs were more stereotypedthan female songs that song. Thus, a larger HAD indicates a sup- for all the song types examined. Therefore, I pressed harmonic component and a pure tonal concluded males consistently repeat syllables in quality. A smaller HAD indicates a louder sec- a more stereotypedmanner than females do re- ond harmonic component resulting in a nasal gardless of song types (Fig. 3A). sound quality. HARMONIC AMPLITUDE DIFFERENCE (HAD) STATISTICAL ANALYSIS Spectral structure differed considerably between The question addressedin the analyses is wheth- the songs of the two sexes. HAD was signifi- er the acoustic parameters measured differed re- cantly higher in male songs than in female liably between the two sexes. Because song songs, indicating that female songs contain sig- types are acoustically diverse, the measured pa- nificantly louder second harmonics than male rameters could differ among song types as well songs (Table 1). Although HAD differed among as between the sexes. In an effort to tease apart different song types (F7,2o = 19.7, P < O.OOl), the influences of song type and sex on the mea- male songs expressedless of second harmonics sured variables, I used a two-way analysis of than female songs did within each song type, variance (ANOVA; sex and song type were fac- and this trend was consistent across different SEXUALLY DIMORPHIC LEARNED SONG 507

A. a

syllable 1 2a 2b 3a 3b

phrase 1 phrase 2

0 5 set B. Qz-80 -----

;E; -100 Harmonic Amplitude Difference s s e -120 A-I

Frequency (kHz) FIGURE 1. Method used to measure acoustic parameters of male and female songs. (A) A spectrogram illus- trating structure of a song consisting of two phrases, each containing repeated notes/syllables. Notes contained in phrase 1 are not considered to be two-noted syllables because they are separated by less than 8 msec (see Methods). For Syllable Repetition Accuracy (SRA) measures, cross correlation coefficients between syllable 1 and 2, 2 and 3 for phrase 1, and coefficients between syllable la and 2a, 2a and 3a, lb and 2b, 2b and 3b for phrase 2 were calculated, and the average coefficient for each phrase was computed. Then, these values were averaged for each song. Shaded segments marked with xf indicate sections of sounds used to generate amplitude spectra to estimate Harmonic Amplitude Difference (HAD) of a song in B. (B) Mean HAD were calculated based on three spectra generated from time frames such as those indicated by shaded area with xf in (A). The amplitude difference of the first peak (the fundamental frequency, marked with X) and the second peak (the second harmonic overtone, marked with 2X) was measured.Note that portions of songs sampled for spectra generation had a fundamental frequency below 5 kHz. For each pair of male and female songs used in the analysis, homologous syllable pairs were chosen for SRA calculation, and similar loci were chosen for spectra calculation.

song types (interaction effect of sex and song ple (n = 36). The distribution of the data set type; F,,20= 0.8, P > 0.5). Therefore, I conclude required a quadratic regression model. The re- female songs contain harmonics that are consis- sult showed that, for both sexes, the fundamental tently louder relative to the fundamental than frequency contributed significantly to the varia- male songs do regardless of song type (Fig. 3B). tion in HADs (R2 = 0.67, linear term: F1,32= The observed variation in harmonic content 35.3, P < 0.001, quadratic term : F,,32= 6.8, P among different song types could be ascribed to < 0.02, Fig. 4). In this analysis, HAD also dif- differences in the frequency content of song fered significantly between the songs of the two types. To examine if harmonic structure is influ- sexes (F,,32= 20.7, P < O.OOl), but there was enced by the fundamental frequency (Fig. 1B no sex difference in fundamental frequency (un- frequency point indicated by X) as well as by paired t-test, t,, = -0.01, P > 0.5). Therefore, sex, I carried out an analysis of covariance (AN- I conclude that the harmonic amplitude differ- COVA) with sex as a factor and fundamental ence increases with the fundamental frequency frequency as a covariate. For this analysis, one asymptotically, and the HAD is always larger in spectrum from each song was selected as a sam- male songs than in female songs. 508 AYAKO YAMAGUCHI

TABLE 1. The results of an ANOVA examining sex *la Tib differences in measuredacoustic parameters. Two-way 7- ANOVA F-statistic (df = 1, 20) and P-value of acoustic parametersfor factor “sex.” Asterisks indicate pa- 6- rameters that differed between the two sexes. 5 - 4- PZCUll&%S F P-Vkle 3- Frequency measurements Maximum frequency 1.3 0.27 Minimum frequency 1.1 0.30 Dominant frequency 4.0 0.06 Frequency range 0.5 0.48 Syllable Repetition Accuracy 10.3 <0.01* d Harmonic Amplitude Difference 26.6 <0.001* I !

DISCUSSION The results show that, despite their similarity to the human ear, the songs of cardinals are consistently sexually dimorphic in syllable stereo- o- - typy and tonal quality. Male and female play- II back songs used in my previous field study also I Pd Pd I were sexually dimorphic with respect to these sex two parameters (Yamaguchi, in press). Thus, FIGURE 2. Cell plots showing mean 2 SE of four these differences between the songs of the two frequencymeasures of male and female cardinal songs. sexes may be used by cardinals to distinguish a. maximum, b. minimum, c. frequency range, d. dominant frequency. None of the frequency measuresdif- the sex of the singer. fered between the two sexes. How do the differences between the sexes develop? One possibility is that the sex-specific traits of songs are acquired by young birds through imitating only adult songs of their own sex. However, the possibility of a sex-specific song tradition can be ruled out on the basis of results from a separate study of song development in cardinals (Yamaguchi 1996). I found that when juvenile cardinals are given a choice of songs produced by males and females, they learn songs from both sexesequally well. More- over, songs developed by males and females were either masculinized or feminized with respect to syllable stereotypyregardless of the sex of the model songs that were imitated. Thus, the sex dimorphism of cardinal songsis best viewed as an expressionof some inherent morphological I -10' A and/or physiological differences between the 1 2 3 4 5 6 7 6 two sexes, not as the result of sex-specific song Song Types learning. What kind of physiological and/or FIGURE 3. (A) Scatter plot of Syllable Repetition morphological differences might give rise to the Accuracy t SE of the two sexes separatedby song observed acoustic dimorphism? types. SRAs are similar among different song types, One of the sexually dimorphic traits revealed and males consistentlyhave significantly higher SRAs by quantitative analysis, the difference in degree than females do. (B) Scatter plot of Harmonic Ampli- tude Difference 5 SE of the two sexes separatedby of syllable stereotypy, also was evident in post- song types. HADs vary significantly among song hoc spectrographicobservations. The frequency types, but female songs contain significantly larger modulation of spectrographictraces tends to be HADs than male songsdo regardlessof song types. SEXUALLY DIMORPHIC LEARNED SONG 509

50 - Z- 9 $ 40- s g 30- AQ ii $ 20- a I’a o Female 2 lo- l Male .o El 6 E ’ &i I 0 -10 I * I . 8. I . 1 - 1 2 3 4 5 Fundamental Frequency (kHz) FIGURE 4. The relationship between the fundamental frequency, sex, and the harmonic amplitude difference (HAD). The harmonic amplitudedifference increaseasymptotically along with increasedfundamental frequency, and male songs always have larger HADs than female songsdo within the 5 kHz analysisrange.

less smooth in female than in male syllables, and male Song Sparrows Melospiza melodia contain the shapes of the repeated female syllables are louder harmonics than crystallized adult songs, more irregular in comparison to those of repeat- and that the suppression of harmonic amplitude ed male syllables. Typically, in male songbirds, is a skill that emerges during the course of vocal the vocal output progresses from a less stereo- motor development. It is likely that male cardi- typed (plastic song) to a highly stereotyped form nals follow a similar developmental trajectory, (crystallized song) during the sensorimotor enhancing the tonal quality of songs during the phase of song development (Marler and Peters sensorimotor phase. The relative amplitude of 1982, Podos et al. 1995). Male cardinals also harmonics contained in adult female songs re- show the characteristic increase in vocal stereo- sembles that in plastic songs of male cardinals. typy prior to song crystallization. The lower de- Therefore, the acoustic morphology of a female gree of syllable stereotypy in songs of adult fe- song may be viewed as a neotenic form of the male cardinals suggests that perhaps females crystallized male song. never fully crystallize their songs. Sex differences in the level of syllable stereo- The other sexually dimorphic trait I found in typy and harmonic amplitude may be ascribable this study is the relative amplitude of harmonics; to hormonal differences in the two sexes. Pre- female songs contain relatively louder harmon- vious studies have reported that testosterone is ics than male songs do. Nowicki and his col- necessary for song crystallization in male song- leagues have shown that in songbirds, harmonics birds (Marler et al. 1987, Marler et al. 1989). In are generated at the syringeal sound source, and a separate study of the endocrinological basis of are efficiently filtered out by the coordinated ac- song development in cardinals, I found that fe- tivity of the syrinx and vocal tract movement; males maintain low basal levels of testosterone the result is a concentration of energy in the fun- throughout song development, whereas males damental frequency that gives songs a pure-tonal have elevated levels coinciding both with the quality in many species of songbirds (Nowicki sensorimotor period of song learning and with and Capranica 1986, Nowicki 1987). Female the song production period in adulthood (Ya- cardinals are sometimes observed to sing with- maguchi 1996). Thus, differences in the circu- out opening their beak in contrast to males that lating levels of testosterone between the two sing with their beaks flared (Yamaguchi, pers. sexes may account for the sex difference in syl- observ.). This difference in singing posture lable stereotypy and harmonic amplitude. If this probably affects the amplitude of harmonics in hypothesis is correct, then I predict that admin- songs of the two sexes. Recently, Podos et al. istration of testosterone to a female in the later (1995) showed that plastic songs of immature stages of song development should masculinize 510 AYAKO YAMAGUCHI her songs by increasing syllable stereotypy and one at later stagesof song development. The di- suppressingharmonic amplitude. The target sites morphism is sufficiently stable and predictable of testosteroneare likely to be both peripheral that it probably explains acousticsex recognition and central. Testosterone is shown to act as a as exhibited by cardinals in a previous study trophic factor and increase the weight of syrinx (Yamaguchi, in press). (Luine et al. 1980). It is conceivable that the sex difference in the syringeal muscle mass at the ACKNOWLEDGMENTS time of song production leads to difference in I thank Peter Marler for his helpful guidance,support, sound quality. In addition, testosteronemay act advice, and comments on this research and on the on the central nervous system and elicit changes manuscript.I thank Doug Nelson and Kim Beeman for in motor patterns including posture assumed their advice on experimental design and sound analy- while singing and the coordinated movement of sis, and two anonymousreviewers for comments.This researchwas supportedby the Chapman Fund of the trachea such that harmonic structureof the songs American Museum of Natural History, Sigma Xi, and become different in the two sexes. Jastro Shields Grants from the Universitv of Califor- One of the acoustical differences of songsde- nia, Davis. scribed here, the harmonic amplitude difference, may be related to the functional differences of LITERATURE CITED songs in the two sexes. Male cardinals, as in BALLINTIJN,M. R., AND C. TENCATE. 1997. Sex dif- many other species of oscines, sing mainly to ferences in the vocalizations and syrinx of the defend their territories and to attract females Collared Dove (Streptopelia decaocto). Auk 114: (Laskey 1944). Female cardinals, in contrast, 22-39. BENT,A. C. 1968. Life histories of North American use their songs for intra-pair communication cardinals, grosbeaks, bunting, towhees, finches, within a territory (Ritchison 1988, Halkin 1997) sparrows,and allies. Smithson. Inst. Press, Wash- and in female-female agonistic interactions ington, DC. when a strange female intrudes into a territory CLARK, C. W., l? MARLER,AND K. BEEMAN. 1987. already occupied by a female (Ritchison 1988, Quantitative analysis of animal vocal phonology: an application to swamp sparrow song. Ethology Yamaguchi, pers. observ.). Thus, the male’s 76:101-l 15. song is a long range signal that must be loud to Drrrus, W. E J., AND R. E. LEMON. 1969. Effects of reach its audience, whereas the female’s song is song tutoring and acoustic isolation on the song a short range signal that can be soft. Peters and repertoires of cardinals. Anim. Behav. 17:523- Nowicki (1996) suggestthat it may be easier for 533. ENGINEERINGDESIGN. 1987. Signal software manual. birds to produce loud sounds if the sound is a Engineering Design, Belmont, MA. pure tone. The prediction derived from this hy- FAY, R. R. 1988. Hearing in vertebrates:a psycho- pothesis, a tight coupling between higher sound physics databook.Hill-Fay Associates,Winnetka, amplitude and pure-tonal quality of sound, is IL. prevalent in birdsong in general. Males of mo- GAHR,M., AND H-R G~JITINGER.1986. Functional as- pects of singing in male and female Urueginthus nogamous territorial speciessuch as Song Spar- bengalus (Estrildidae). Ethology 72: 123-13 1. rows and cardinals (i.e., long range signalers) HALKIN,S. L. 1990. Singing from the nest: intrapair tend to produce pure-tonal songs,whereas males communication in cardinals. Ph.D. diss., Univ. of nonterritorial breeders (i.e., short range sig- Wisconsin, Madison, WI. HALK~N,S. L. 1997. Nest-vicinity song exchanges nalers) such as Zebra Finches (Taeniopygia gut- may coordinate biparental care of Northern tutu) produce songs with significant harmonic Cardinals. Anim. Behav. 54:89-198. overtones. Therefore, the differential amplitude KONISHI,M. 1963. The role of auditory feedback in requirement of male and female songsmay have the vocal behavior of the domestic fowl. Z. Tierp led to the evolution of sexual dimorphism in har- sychol. 20:329-367. KROODSMA,D. E. 1982. Learning and the ontogeny monic amplitude in cardinals. of sound signals in birds, p. l-23. In D. E. I have demonstrated the presence of predict- Kroodsma and E. H. Miller leds.1.Acoustic com- able and reliable sexual dimorphism in a learned munication in birds. Vol. 2. Academic Press,New vocalization. I conclude that acoustic dimor- York. phism of cardinal song can be viewed as a dif- LASKEY,A. R. 1944. A study of the cardinal in Ten- nessee.Wilson Bull. 56:27-44. ference in the completeness of song develop- LEMON,R. E., AND C. CHATFIELD.1971. Organization ment in the two sexes, and it may reflect sex of song in cardinals.Anim. Behav. 19:1-17. differences in the circulating levels of testoster- LEMON,R. E., AND D. M. Scorr. 1966. On the de- SEXUALLY DIMORPHIC LEARNED SONG 511

velopment of song in young cardinals. Can. J. NOWICKI,S., AND R. R. CAPRANICA.1986. Bilateral Zool. 44:191-197. syringeal interactionin vocal productionof an os- LUINE, V., E NOTTEBOHM,C. HARDING,AND B. S. tine bird sound. Science 23 1:1297-1299. MCEWEN. 1980. Androgen affects cholinergicen- PETERS,S., AND S. NOWICIU. 1996. Development of zymes in syringeal motor neurons and muscle. tonal quality in birdsong: further evidence from Brain Res. 192:89-107. Song Sparrows.Ethology 102:323-335. MARLER,I?, ANDS. PETERS.1982. Subsongand plastic PODOS,J., J. K. SHERER, S. PETERS,AND S. NOWICKI. song: their role in the vocal learning process, p. 1995. Ontogeny of vocal tract movementsduring 25-50. In D. E. Kroodsma and E. H. Miller [eds.], song production in song sparrows.Anim. Behav. Acoustic communication in birds. Vol. 2. Aca- 50: 1287-1296. demic Press, New York. RITCHISON,G. 1988. The singing behavior of female MARLER,P, S. PETERS,G. I? BALL, A. M. DUFTYJR., Northern Cardinals. Condor 88: 156-l 59. AND J. C. WINGFIELD.1989. The role of sex ster- TAOKA,M., AND H. OKUMURA. 1990. Sexual differ- oids in the acquisitionand productionof birdsong. ences in flight calls and the cue for vocal sex rec- Nature 336:770-771. ognition of Swinhoe’s Storm-petrel. Condor 92: MARLER,P., S. PETERS,AND J. C. WINGFIELD.1987. 571-575. Correlation between song acquisition, song pro- YAMAGUCHI,A. 1996. Female bird song: function, duction, and plasma levels of testosteroneand es- physiology, and developmentin the Northern Car- tradiol in sparrows.J. Neurobiol. 186:531-548. dinal. Ph.D. diss., Univ. California, Davis, CA. NOWICKI,S. 1987. Vocal tract resonancesin oscine YAMAGUCHI,A. In press. Can a sexually dimorphic bird sound production: evidence from bird-songs learned birdsong be used for male-female recog- in a helium atmosphere.Nature 325:53-55. nition? Behaviour.